• Monolith
  • Posts
  • Instagram, a comment waterfall, and a single server

Instagram, a comment waterfall, and a single server

Instagram’s feed at launch

6 min read

With the recent Threads hype cycle, I figured it would be a good time to cover Instagram.

Brief history: Instagram’s origin story is about as close to a SV fairytale as one could tell. It was created by Kevin Systrom and Mike Krieger. In 2009 Kevin was building iPhone apps on the side. At the time check-in apps were all the rage. FourSquare was the leader. He thought he could do better so he built Burbn. Burbn included the ability to share photos alongside checkins, something FourSquare did not support at the time. In March 2010 Kevin showed up at an investor meetup and met Steve Andersen. They met for coffee, and Steve committed to $50,000 under one condition: Kevin find a co-founder. Kevin closed a $500,000 seed with $250,000 from Andreessen Horowitz, and Mike Kreiger, a college friend, agreed to join as a co-founder.

Through the summer of 2010 they only had a total of 100 users. They had a “coming to Jesus” moment about the failure. They looked into what users liked from the app and made a list of 3 things. Photo sharing was at the top. They were burnt out so Kevin decided to take a trip to Mexico with his wife (9:38):

We were both burned out so I was like, “Alright Mike I have to take a break”. So my, now wife, and I, Nicole, went to Mexico. We were working on this idea of photos so Mike was like “okay I’m going to prototype this thing while you’re on break. You come back and then we’ll reconvene.” So Mike started prototyping this app. As I’m away, I’m walking down the beach with Nicole in Mexico. We were in this little surfing town called Todos Santos. I’m like “Nicole, I think we’re going to focus on photos.” And she goes “Oh yeah, that sounds like a great idea. I love all the photo posts on Burbn.” But she goes “I don’t think I’m going to post that much.” And I go “well why not?”. And she goes “Well my photos aren’t that good.” And I said “well they’re great”. And she goes “they’re not as good as your friend Greg.” And I go “Greg uses a bunch of filter apps to make them look good.” And she goes “Well you should probably add filters.”

This was the key insight. They needed to make users feel like their photos are worthy of sharing. Kevin goes back to the hotel room, prototypes the first filter, and posts a picture with the filter he built.

The test photo Kevin posted with the new filter

Instagram launches on the Apple App Store on October 6th, 2010. Within 24 hours they had 25,000 downloads and the rest is history.

They hit 1M users in 3 months, and 2M users six weeks later. They raised a $7M Series A in February led by Benchmark.


Comments and likes were not collapsed in the main feed

Every single comment and like was shown. Mike talking about this:

We had no aggregation logic for likes or comments because we had only tested it with our group of 100 TestFlight users and no post ever got more than 20 likes. You could imagine that when we launched and it actually took off that it did not go well. The first week if you had 1,000 comments on your post, it would just have 1,000 comments. We shipped an update pretty quickly, but you know, it’s funny, in some ways it’s clown town but if we had spent all this time worrying about it maybe we would have slowed down launch.

They cropped photos on device to only upload the dimensions they needed

Mike talking about this as he walks through the original Instagram demo video:

The reason uploads were so fast in Instagram at the time was that we would upload exactly, this was pre-retina display, 306 pixels. You might ask why? Well the screen was 320 wide, and we had a 6 pixel inset on either side so we uploaded exactly what we needed to get that uploaded at full res on the one device that we supported which was the iPhone.

The initial filters looped through every pixel and applied filters

Mike again talking about this with the demo video:

At the time these were written in C, so these were going pixel by pixel and applying the color transformer, and then overlaying a border. In 2011, we rewrote the filter stack to use OpenGL, and it was much faster and we could do fancier things but this was the original version.

Instagram ran on a single computer in a colo in LA

As you can imagine, they had some issues scaling. Kevin discussing the initial launch (12:28):

Before you knew it we actually had overloaded our system, and it was a very small, naive system. It was a single computer in a colocation space somewhere in LA. There was nothing more exhilarating than seeing these people stream in and nothing more crushing than seeing people posting on Twitter or on their blogs and saying “Oh another startup that doesn’t know how to scale. Like so clowny”. We were both at that point running on zero sleep for like two days, devastated. I was like “This is it. We built this great thing, and we totally messed it up.”

My take

They focused incessantly on the user

Two examples stuck with me. The first is the original pivot and the idea for filters. Instagram was special because it made their users special. They learned this from talking with users. Second, in the Instagram demo video, Mike talks repeatedly about how they focused on the UX of Instagram. Comments were rendered optimistically. Photos were uploaded in the background as the user typed out the caption and added a location. They only uploaded the exact number of pixels they needed.

Timing was a big part of their success

They believed that mobile was going to change what data people were sharing about themselves. Their first thought was that this would be their location, hence Burbn. Obviously, this turned out to be photos.

Starting at the beginning of the mobile revolution also meant that phone cameras were not as good. This allowed Instagram to be differentiated with filters. The use of filters and quality of built-in cameras are inversely correlated. Instagram likely wouldn’t have survived if it had launched in a world with high quality cameras.

One of their core principles was “do the simple thing first”

Mike talks about this with the comments example above. This is something I’ve tried to become better at in my engineering career too. It’s antithetical to much of what I learned in school.